headers
Andrey Borzenkov | 1 Jan 2007 20:44
Picon

2.6.20 regression: suspend to disk no more works


In *the same* configuration STD now fails with "Cannot find swap device". The 
reason is changes in kernel/power/swap.c. In 2.6.19 it did not require valid 
swsusp_resume_device at all - it took first available swap device and saved 
image. Later during resume swsusp_resume_device was set either by command 
line or sysfs and everything worked nicely.

Now swsusp_swap_check() unfortunately checks for swsusp_resume_device at 
*suspend* time:

        res = swap_type_of(swsusp_resume_device, swsusp_resume_block);
        if (res < 0)
                return res;

        root_swap = res;
        resume_bdev = open_by_devnum(swsusp_resume_device, FMODE_WRITE);
        if (IS_ERR(resume_bdev))
                return PTR_ERR(resume_bdev);

but in case of modular driver for swap device this is likely to be undefined. 
This is as of 2.6.20-rc3.

I already have seen these reports. While 'echo a:b > /sys/power/resume' before 
suspend is a workaround, this still breaks perfectly valid setup that worked 
before. Also 'echo a:b > /sys/power/resume' is actually wrong - we are not 
going to resume at this point; but there is no way to just tell kernel "use 
this device for next STD" ... also the error message is misleading, it should 
complaint "no resume device found". Swap is there all right.

-andrey
(Continue reading)

Permalink | Reply | 0 comments |
headers
Rafael J. Wysocki | 2 Jan 2007 00:28
Picon
Gravatar

Re: 2.6.20 regression: suspend to disk no more works

Hi,

On Monday, 1 January 2007 20:44, Andrey Borzenkov wrote:
> In *the same* configuration STD now fails with "Cannot find swap device". The
> reason is changes in kernel/power/swap.c. In 2.6.19 it did not require valid
> swsusp_resume_device at all - it took first available swap device and saved
> image. Later during resume swsusp_resume_device was set either by command
> line or sysfs and everything worked nicely.
> 
> Now swsusp_swap_check() unfortunately checks for swsusp_resume_device at
> *suspend* time:
> 
>         res = swap_type_of(swsusp_resume_device, swsusp_resume_block);
>         if (res < 0)
>                 return res;
> 
>         root_swap = res;
>         resume_bdev = open_by_devnum(swsusp_resume_device, FMODE_WRITE);
>         if (IS_ERR(resume_bdev))
>                 return PTR_ERR(resume_bdev);
> 
> but in case of modular driver for swap device this is likely to be undefined.
> This is as of 2.6.20-rc3.

Actually, if you look at the 2.6.19 code, the call to swap_type_of() is there
in swsusp_swap_check() too.

The problem is with open_by_devnum(), I think, which obviously cannot succeed
if swsusp_resume_device is not set.  I think we should return resume_bdev from
swsusp_swap_check() like in the appended patch (untested).  Pavel?
(Continue reading)

Permalink | Reply | 0 comments |
headers
Andrey Borzenkov | 2 Jan 2007 09:38
Picon

Re: 2.6.20 regression: suspend to disk no more works


On Tuesday 02 January 2007 02:28, Rafael J. Wysocki wrote:
> Hi,
>
> On Monday, 1 January 2007 20:44, Andrey Borzenkov wrote:
> > In *the same* configuration STD now fails with "Cannot find swap device".
> > The reason is changes in kernel/power/swap.c. In 2.6.19 it did not
> > require valid swsusp_resume_device at all - it took first available swap
> > device and saved image. Later during resume swsusp_resume_device was set
> > either by command line or sysfs and everything worked nicely.
> >
> > Now swsusp_swap_check() unfortunately checks for swsusp_resume_device at
> > *suspend* time:
> >
> >         res = swap_type_of(swsusp_resume_device, swsusp_resume_block);
> >         if (res < 0)
> >                 return res;
> >
> >         root_swap = res;
> >         resume_bdev = open_by_devnum(swsusp_resume_device, FMODE_WRITE);
> >         if (IS_ERR(resume_bdev))
> >                 return PTR_ERR(resume_bdev);
> >
> > but in case of modular driver for swap device this is likely to be
> > undefined. This is as of 2.6.20-rc3.
>
> Actually, if you look at the 2.6.19 code, the call to swap_type_of() is
> there in swsusp_swap_check() too.
>
> The problem is with open_by_devnum(), I think, which obviously cannot
(Continue reading)

Permalink | Reply | 0 comments |
headers
Pavel Machek | 2 Jan 2007 11:32
Picon

Re: 2.6.20 regression: suspend to disk no more works

Hi!

> > While 'echo a:b > /sys/power/resume' before 
> > suspend is a workaround, this still breaks perfectly valid setup that worked
> > before. Also 'echo a:b > /sys/power/resume' is actually wrong - we are not
> > going to resume at this point; but there is no way to just tell kernel "use
> > this device for next STD" ... also the error message is misleading, it should
> > complaint "no resume device found". Swap is there all right.
> 
> Thanks for the report.

It fixes it for too people, I guess that's ACK... certainly for -mm,
probably for 2.6.20, too.
								Pavel

--

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
Permalink | Reply | 0 comments |
headers
Adrian Bunk | 2 Jan 2007 20:24
Picon

2.6.20-rc3: known regressions with patches available (part 1)

This email lists some known regressions in 2.6.20-rc3 compared to 2.6.19
with patches available.

If you find your name in the Cc header, you are either submitter of one
of the bugs, maintainer of an affectected subsystem or driver, a patch
of you caused a breakage or I'm considering you in any other way possibly
involved with one or more of these issues.

Due to the huge amount of recipients, please trim the Cc when answering.

Subject    : suspend to disk no longer works
References : http://lkml.org/lkml/2007/1/1/72
             http://lkml.org/lkml/2006/12/31/13
Submitter  : Andrey Borzenkov <arvidjaar <at> mail.ru>
             Robert Hancock <hancockr <at> shaw.ca>
Handled-By : Rafael J. Wysocki <rjw <at> sisk.pl>
Patch      : http://lkml.org/lkml/2007/1/1/117
Status     : patch available

Subject    : "i386: Relocatable kernel support" causes instant reboot
References : http://lkml.org/lkml/2006/12/20/114
Submitter  : Jean Delvare <khali <at> linux-fr.org>
Caused-By  : Eric W. Biederman <ebiederm <at> xmission.com>
             commit 968de4f02621db35b8ae5239c8cfc6664fb872d8
Handled-By : Vivek Goyal <vgoyal <at> in.ibm.com>
Patch      : http://lkml.org/lkml/2007/1/2/9
Status     : patch available

Subject    : bluetooth oopses because of multiple kobject_add()
References : http://lkml.org/lkml/2007/1/2/101
(Continue reading)

Permalink | Reply | 0 comments |
headers
Jens Axboe | 2 Jan 2007 20:26
Picon
Favicon

Re: 2.6.20-rc3: known regressions with patches available (part 1)

On Tue, Jan 02 2007, Adrian Bunk wrote:
> Subject    : CFQ disk throughput halved
> References : http://lkml.org/lkml/2007/01/1/104
> Submitter  : Rene Herman <rene.herman <at> gmail.com>
>              Mark Lord <lkml <at> rtr.ca>
> Caused-By  : Jens Axboe <jens.axboe <at> oracle.com>
>              commit 719d34027e1a186e46a3952e8a24bf91ecc33837
> Handled-By : Jens Axboe <jens.axboe <at> oracle.com>
> Patch      : http://lkml.org/lkml/2007/1/2/75
> Status     : patch available

Patch is already merged in -git.

--

-- 
Jens Axboe
Permalink | Reply | 0 comments |
headers
Adrian Bunk | 2 Jan 2007 20:34
Picon

Re: 2.6.20-rc3: known regressions with patches available (part 1)

On Tue, Jan 02, 2007 at 08:26:52PM +0100, Jens Axboe wrote:
> On Tue, Jan 02 2007, Adrian Bunk wrote:
> > Subject    : CFQ disk throughput halved
> > References : http://lkml.org/lkml/2007/01/1/104
> > Submitter  : Rene Herman <rene.herman <at> gmail.com>
> >              Mark Lord <lkml <at> rtr.ca>
> > Caused-By  : Jens Axboe <jens.axboe <at> oracle.com>
> >              commit 719d34027e1a186e46a3952e8a24bf91ecc33837
> > Handled-By : Jens Axboe <jens.axboe <at> oracle.com>
> > Patch      : http://lkml.org/lkml/2007/1/2/75
> > Status     : patch available
> 
> Patch is already merged in -git.

Thanks for this information, I missed this (as well as the merged SATA 
fix) since it isn't yet at the git mirrors.

> Jens Axboe

cu
Adrian

--

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed
(Continue reading)

Permalink | Reply | 0 comments |
headers
Rene Herman | 2 Jan 2007 21:45
Picon

Re: 2.6.20-rc3: known regressions with patches available (part 1)

Adrian Bunk wrote:

> On Tue, Jan 02, 2007 at 08:26:52PM +0100, Jens Axboe wrote:

>> Patch is already merged in -git.
> 
> Thanks for this information, I missed this (as well as the merged SATA 
> fix) since it isn't yet at the git mirrors.

What's "-git" here? I just now pulled

git://git2.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git

and it's not there.

Otherwise, patch confirmed to work by me as well.

Cheers,
Rene
Permalink | Reply | 0 comments |
headers
Jens Axboe | 2 Jan 2007 21:51
Picon
Favicon

Re: 2.6.20-rc3: known regressions with patches available (part 1)

On Tue, Jan 02 2007, Rene Herman wrote:
> Adrian Bunk wrote:
> 
> >On Tue, Jan 02, 2007 at 08:26:52PM +0100, Jens Axboe wrote:
> 
> >>Patch is already merged in -git.
> >
> >Thanks for this information, I missed this (as well as the merged SATA 
> >fix) since it isn't yet at the git mirrors.
> 
> What's "-git" here? I just now pulled
> 
> git://git2.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
> 
> and it's not there.
> 
> Otherwise, patch confirmed to work by me as well.

Perhaps not mirrored out yet, I pulled it from master/hera some hours
ago though. The top of Linus' tree is
ec8acb6904fabb8e741f741ec99bb1c18f2b3dee, which is the commit with that
patch.

--

-- 
Jens Axboe
Permalink | Reply | 0 comments |
headers
Adrian Bunk | 3 Jan 2007 22:04
Picon

2.6.20-rc3: known regressions with patches (v2)

This email lists some known regressions in 2.6.20-rc3 compared to 2.6.19
with patches available.

If you find your name in the Cc header, you are either submitter of one
of the bugs, maintainer of an affectected subsystem or driver, a patch
of you caused a breakage or I'm considering you in any other way possibly
involved with one or more of these issues.

Due to the huge amount of recipients, please trim the Cc when answering.

Subject    : suspend to disk no longer works
References : http://lkml.org/lkml/2007/1/1/72
             http://lkml.org/lkml/2006/12/31/13
Submitter  : Andrey Borzenkov <arvidjaar <at> mail.ru>
             Robert Hancock <hancockr <at> shaw.ca>
Handled-By : Rafael J. Wysocki <rjw <at> sisk.pl>
Patch      : http://lkml.org/lkml/2007/1/1/117
Status     : patch available

Subject    : BUG at drivers/scsi/scsi_lib.c:1118 by "pktsetup dvd /dev/sr0"
References : http://bugzilla.kernel.org/show_bug.cgi?id=7667
Submitter  : Laurent Riffard <laurent.riffard <at> free.fr>
Caused-By  : Christoph Hellwig <hch <at> lst.de>
             commit 3b00315799d78f76531b71435fbc2643cd71ae4c
Handled-By : Christoph Hellwig <hch <at> lst.de>
Patch      : http://bugzilla.kernel.org/show_bug.cgi?id=7667
Status     : patch available

Subject    : forcedeth.c 0.59: problem with sideband managment
References : http://bugzilla.kernel.org/show_bug.cgi?id=7684
(Continue reading)

Permalink | Reply | 0 comments |

Gmane